Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2610.16270

Continuity of certain quantum channel divergences

Morgan Jones

quant-phSubmitted Oct 12, 2026• Announced Oct 13, 2026• 0 citations• Revised Oct 21, 2026 (v2)
v1 · Oct 12, 2026v2 · Oct 21, 2026
arXivHTMLPDF

Abstract

The geometric channel fidelity, the geometric Rényi divergences and the Belavkin--Staszewski relative entropy give converse bounds for capacities, discrimination and estimation. Channels are known only approximately, yet two pairs of channels with equal sums can be arbitrarily close in the diamond norm and have fidelities $0$ and $1$. The remedy is the Arveson--Radon--Nikodym derivative $T$ of one map with respect to the sum $Ω$, whose transpose is the environment effect selecting that map in a Stinespring dilation of $Ω$. In $T$, these quantities and all maximal $f$-divergences of operator convex $f$ are continuous, with explicit dimension-free moduli; for two pairs of channels with the same sum the fidelities differ by at most $2\|T_1-T_2\|_\infty^{1/2}$, which gives Hölder continuity in the diamond norm on such splittings. The fidelity is continuous at a pair of completely positive maps exactly when the Choi matrix of the sum is invertible or the fidelity vanishes: noisy channels of full Kraus rank are continuity points, unitary gates are not.

AI Summary

The paper studies continuity of geometric channel fidelity, geometric Rényi quantities and maximal $f$-divergences under perturbations of completely positive maps. Diamond-norm closeness alone gives no uniform control: arbitrarily close pairs of replacement channels with equal sums can have geometric fidelities $0$ and $1$, reflecting the known discontinuity of Matsumoto fidelity on distinct pure states. For a fixed sum $\Omega=\Phi+\Psi$, it uses the Arveson–Radon–Nikodym derivative $0\leq T\leq I$ of $\Phi$ relative to $\Omega$. Its transpose is the environment effect selecting one branch. Kubo–Ando means and operator perspectives reduce the divergences to functional calculus in $T$, followed by a Kraus-dependent partial trace and an extremal eigenvalue. In particular, $|\widehat F_1-\widehat F_2|\leq\|\Omega^*(I)\|_\infty\|T_1-T_2\|_\infty^{1/2}$, with prefactor $2$ for pairs of channels. Integral representations yield dimension-free moduli for all means and operator-convex maximal $f$-divergences. Fixed-sum diamond-norm bounds additionally depend on a conditioning parameter of $\Omega$; finite quantitative bounds for unbounded divergences require max-relative-entropy control. A dilation construction characterises continuity on pairs of completely positive maps: geometric fidelity is continuous exactly when $J(\Phi+\Psi)$ is invertible or the fidelity is zero. Analogues cover Rényi quasi-entropies and means; bounded maximal $f$-divergences require an additional rank condition for necessity. The paper also treats changing sums and supports. Limitations include unoptimised constants, an unresolved extension of the full dichotomy to channel-valued perturbations, and no quantitative application to a specific operational task.

No sign of AI-generated text

Error found: The abstract does not qualify its blanket promise of explicit continuity moduli for all maximal operator-convex f-divergences: finite-difference bounds for unbounded examples require spectral restrictions, whereas the unrestricted continuity asserted in the body is extended-valued.

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Related papers

Continuity of Regularized Channel Rényi Divergences

Jinzhao Wang, Yuxiang Yang

We prove that the regularized, stabilized sandwiched Rényi divergence of finite-dimensional quantum channels converges to their regularized relative entropy as the Rényi order tends to one. The key tool is the channel hockey-stick divergence: Gour's Stinespring approximation bound and a Schatten norm estimate amplify an asymptotic bound below one into exponential decay at higher threshold rates. For channel pairs with finite max-relative entropy, known operational connections then give exponential strong converses for parallel and adaptive discrimination, a sharp zero--one testing law, and the subchannel asymptotic equipartition property.

quant-pharXiv:2609.28635•Sep 25, 2026•1 citation

INFORMATION DIVERGENCE FOR QUANTUM CHANNELS

S.J. HAMMERSLEY, V. P. Belavkin

Many definitions of relative information (often called relative entropy) have been proposed in order to characterise the ‘distance’ as information divergence between states. One such class of relative informations are the quasi-entropies proposed by Petz in 66 (of which the standard Araki-Umegaki form, 53 , is a particular example). This paper axiomatises the idea of relative information and then proceeds to build a large class of examples that satisfy these axioms. By using the Radon-Nikodym derivative for completely positive maps as an ‘operational density’ for channels, along with a theorem by Belavkin and Ohya, 57 , for taking the supremum over local couplings these ideas are then lifted beyond the realm of states into that of quantum channels. This method for extending information quantities to channels was pioneered by V.P. Belavkin, with respect to fidelity, 61 (see 60 for an in-depth presentation of the subject).

•Mar 1, 2006•4 citations

Improved Lower Bounds for Learning Quantum Channels in Diamond Distance

Aadil Oufkir, Filippo Girardi

We prove that learning an unknown quantum channel with input dimension $d_A$, output dimension $d_B$, and Choi rank $r$ to diamond distance $\varepsilon$ requires $ Ω\!\left( \frac{d_A d_B r}{\varepsilon \log(d_B r / \varepsilon)} \right)$ channel queries when $d_A= rd_B$, and $Ω\!\left( \frac{d_A d_B r}{\varepsilon^2 \log(d_B r / \varepsilon)} \right)$ channel queries when $d_A\le rd_B/2$. These lower bounds improve upon the best previous $Ω(d_A d_B r)$ bound by introducing explicit, near-optimal $\varepsilon$-dependence. Moreover, when $d_A\le rd_B/2$, the lower bound is optimal up to a logarithmic factor. The proof constructs ensembles of channels that are well separated in diamond norm yet admit Stinespring isometries that are close in operator norm.

quant-pharXiv:2601.04180•Jan 8, 2026•5 citations

Operational distance and fidelity for quantum channels

V. P. Belavkin, Giacomo Mauro D’Ariano, Maxim Raginsky

We define and study a fidelity criterion for quantum channels, which we term the minimax fidelity, through a noncommutative generalization of maximal Hellinger distance between two positive kernels in classical probability theory. Like other known fidelities for quantum channels, the minimax fidelity is well defined for channels between finite-dimensional algebras, but it also applies to a certain class of channels between infinite-dimensional algebras (explicitly, those channels that possess an operator-valued Radon-Nikodym density with respect to the trace in the sense of Belavkin-Staszewski) and induces a metric on the set of quantum channels that is topologically equivalent to the CB-norm distance between channels, precisely in the same way as the Bures metric on the density operators associated with statistical states of quantum-mechanical systems, derived from the well-known fidelity (“generalized transition probability”) of Uhlmann, is topologically equivalent to the trace-norm distance.

arXiv:quant-ph/0408159•Journal of Mathematical Physics•May 20, 2005•52 citations

Quantum gate fidelity in terms of Choi matrices

Nathaniel Johnston, David W. Kribs

We provide new results for computing and comparing the quantum gate fidelity of quantum channels via their Choi matrices. We extend recent work that showed that there exist non-dual pairs of quantum channels with equal gate fidelity by providing an explicit characterization of all such channels. We use our characterization to show that when the dimension is 2 (or 3, under slightly stronger hypotheses), the gate fidelity of two channels is equal if and only if their difference equals the difference of some unital map and its dual—a fact that has been shown to be false when the dimension is 4 or larger. We also present a formula for the minimum gate fidelity of a channel in terms of a well-studied norm on a compression of its Choi matrix. As a consequence, several new ways of bounding and approximating the minimum gate fidelity follow, including a simple semidefinite program to compute it for qubit channels.

arXiv:1102.0948•Journal of Physics A Mathematical and Theoretical•Nov 18, 2011•24 citations

Quantum Channel Stein's Lemma with an Exponential Strong Converse

Minbo Gao, Zhengfeng Ji, Chenghua Liu

We establish the quantum channel Stein lemma at every fixed type-I error tolerance in $(0,1)$ and prove an exponential strong converse under arbitrary adaptive strategies. For any finite-dimensional channel pair, the optimal type-II exponent equals the regularized channel relative entropy and is attained by parallel strategies. When this divergence is finite, every strictly larger type-II rate forces the probability of accepting the null hypothesis to decay exponentially, uniformly over strategies and quantum memories. The proof rests on a uniform Stinespring approximation: tensor powers of the null dilation are approximated by tensor powers of the alternative dilation after applying an auxiliary linear map to the environment. The operator-norm error is exponentially small under a squared-norm bound at any rate above the regularized relative entropy. We construct the approximation from a weak testing bound by iterative rate reduction and a fixed-block tensor expansion with an exact residual. It implies right continuity at order one of the regularized sandwiched Rényi channel divergence, which yields the adaptive converse through the channel chain rule. We also obtain a sharp hockey-stick threshold and a subchannel smoothing asymptotic equipartition property for fixed and subexponentially vanishing diamond-norm errors.

quant-pharXiv:2609.27196•Sep 24, 2026•4 citations
The geometric channel fidelity, the geometric Rényi divergences and the Belavkin--Staszewski relative entropy give converse bounds for capacities, discrimination and estimation. Channels are known only approximately, yet two pairs of channels with equal sums can be arbitrarily close in the diamond norm and have fidelities 000 and 111. The remedy is the Arveson--Radon--Nikodym derivative TTT of one map with respect to the sum ΩΩΩ, whose transpose is the environment effect selecting that map in a Stinespring dilation of ΩΩΩ. In TTT, these quantities and all maximal fff-divergences of operator convex fff are continuous, with explicit dimension-free moduli; for two pairs of channels with the same sum the fidelities differ by at most 2∥T1−T2∥∞1/22\|T_1-T_2\|_\infty^{1/2}2∥T1​−T2​∥∞1/2​, which gives Hölder continuity in the diamond norm on such splittings. The fidelity is continuous at a pair of completely positive maps exactly when the Choi matrix of the sum is invertible or the fidelity vanishes: noisy channels of full Kraus rank are continuity points, unitary gates are not.
The paper studies continuity of geometric channel fidelity, geometric Rényi quantities and maximal fff-divergences under perturbations of completely positive maps. Diamond-norm closeness alone gives no uniform control: arbitrarily close pairs of replacement channels with equal sums can have geometric fidelities 000 and 111, reflecting the known discontinuity of Matsumoto fidelity on distinct pure states.
For a fixed sum Ω=Φ+Ψ\Omega=\Phi+\PsiΩ=Φ+Ψ, it uses the Arveson–Radon–Nikodym derivative 0≤T≤I0\leq T\leq I0≤T≤I of Φ\PhiΦ relative to Ω\OmegaΩ. Its transpose is the environment effect selecting one branch. Kubo–Ando means and operator perspectives reduce the divergences to functional calculus in TTT, followed by a Kraus-dependent partial trace and an extremal eigenvalue. In particular, ∣F^1−F^2∣≤∥Ω∗(I)∥∞∥T1−T2∥∞1/2|\widehat F_1-\widehat F_2|\leq\|\Omega^*(I)\|_\infty\|T_1-T_2\|_\infty^{1/2}∣F1​−F2​∣≤∥Ω∗(I)∥∞​∥T1​−T2​∥∞1/2​, with prefactor 222 for pairs of channels. Integral representations yield dimension-free moduli for all means and operator-convex maximal fff-divergences. Fixed-sum diamond-norm bounds additionally depend on a conditioning parameter of Ω\OmegaΩ; finite quantitative bounds for unbounded divergences require max-relative-entropy control.
A dilation construction characterises continuity on pairs of completely positive maps: geometric fidelity is continuous exactly when J(Φ+Ψ)J(\Phi+\Psi)J(Φ+Ψ) is invertible or the fidelity is zero. Analogues cover Rényi quasi-entropies and means; bounded maximal fff-divergences require an additional rank condition for necessity. The paper also treats changing sums and supports. Limitations include unoptimised constants, an unresolved extension of the full dichotomy to channel-valued perturbations, and no quantitative application to a specific operational task.
We prove that the regularized, stabilized sandwiched Rényi divergence of finite-dimensional quantum channels converges to their regularized relative entropy as the Rényi order tends to one. The key tool is the channel hockey-stick divergence: Gour's Stinespring approximation bound and a Schatten norm estimate amplify an asymptotic bound below one into exponential decay at higher threshold rates. For channel pairs with finite max-relative entropy, known operational connections then give exponential strong converses for parallel and adaptive discrimination, a sharp zero--one testing law, and the subchannel asymptotic equipartition property.
Many definitions of relative information (often called relative entropy) have been proposed in order to characterise the ‘distance’ as information divergence between states. One such class of relative informations are the quasi-entropies proposed by Petz in 66 (of which the standard Araki-Umegaki form, 53 , is a particular example). This paper axiomatises the idea of relative information and then proceeds to build a large class of examples that satisfy these axioms. By using the Radon-Nikodym derivative for completely positive maps as an ‘operational density’ for channels, along with a theorem by Belavkin and Ohya, 57 , for taking the supremum over local couplings these ideas are then lifted beyond the realm of states into that of quantum channels. This method for extending information quantities to channels was pioneered by V.P. Belavkin, with respect to fidelity, 61 (see 60 for an in-depth presentation of the subject).
We prove that learning an unknown quantum channel with input dimension dAd_AdA​, output dimension dBd_BdB​, and Choi rank rrr to diamond distance ε\varepsilonε requires $ Ω\!\left( \frac{d_A d_B r}{\varepsilon \log(d_B r / \varepsilon)} \right)$ channel queries when dA=rdBd_A= rd_BdA​=rdB​, and Ω ⁣(dAdBrε2log⁡(dBr/ε))Ω\!\left( \frac{d_A d_B r}{\varepsilon^2 \log(d_B r / \varepsilon)} \right)Ω(ε2log(dB​r/ε)dA​dB​r​) channel queries when dA≤rdB/2d_A\le rd_B/2dA​≤rdB​/2. These lower bounds improve upon the best previous Ω(dAdBr)Ω(d_A d_B r)Ω(dA​dB​r) bound by introducing explicit, near-optimal ε\varepsilonε-dependence. Moreover, when dA≤rdB/2d_A\le rd_B/2dA​≤rdB​/2, the lower bound is optimal up to a logarithmic factor. The proof constructs ensembles of channels that are well separated in diamond norm yet admit Stinespring isometries that are close in operator norm.
We define and study a fidelity criterion for quantum channels, which we term the minimax fidelity, through a noncommutative generalization of maximal Hellinger distance between two positive kernels in classical probability theory. Like other known fidelities for quantum channels, the minimax fidelity is well defined for channels between finite-dimensional algebras, but it also applies to a certain class of channels between infinite-dimensional algebras (explicitly, those channels that possess an operator-valued Radon-Nikodym density with respect to the trace in the sense of Belavkin-Staszewski) and induces a metric on the set of quantum channels that is topologically equivalent to the CB-norm distance between channels, precisely in the same way as the Bures metric on the density operators associated with statistical states of quantum-mechanical systems, derived from the well-known fidelity (“generalized transition probability”) of Uhlmann, is topologically equivalent to the trace-norm distance.
We provide new results for computing and comparing the quantum gate fidelity of quantum channels via their Choi matrices. We extend recent work that showed that there exist non-dual pairs of quantum channels with equal gate fidelity by providing an explicit characterization of all such channels. We use our characterization to show that when the dimension is 2 (or 3, under slightly stronger hypotheses), the gate fidelity of two channels is equal if and only if their difference equals the difference of some unital map and its dual—a fact that has been shown to be false when the dimension is 4 or larger. We also present a formula for the minimum gate fidelity of a channel in terms of a well-studied norm on a compression of its Choi matrix. As a consequence, several new ways of bounding and approximating the minimum gate fidelity follow, including a simple semidefinite program to compute it for qubit channels.
We establish the quantum channel Stein lemma at every fixed type-I error tolerance in (0,1)(0,1)(0,1) and prove an exponential strong converse under arbitrary adaptive strategies. For any finite-dimensional channel pair, the optimal type-II exponent equals the regularized channel relative entropy and is attained by parallel strategies. When this divergence is finite, every strictly larger type-II rate forces the probability of accepting the null hypothesis to decay exponentially, uniformly over strategies and quantum memories. The proof rests on a uniform Stinespring approximation: tensor powers of the null dilation are approximated by tensor powers of the alternative dilation after applying an auxiliary linear map to the environment. The operator-norm error is exponentially small under a squared-norm bound at any rate above the regularized relative entropy. We construct the approximation from a weak testing bound by iterative rate reduction and a fixed-block tensor expansion with an exact residual. It implies right continuity at order one of the regularized sandwiched Rényi channel divergence, which yields the adaptive converse through the channel chain rule. We also obtain a sharp hockey-stick threshold and a subchannel smoothing asymptotic equipartition property for fixed and subexponentially vanishing diamond-norm errors.